I did a bit more fine tuning today, plus a whole bunch of observations. I found for the current design, nib introduction, speed, quantity, angle of attack to the air stream and dispersion all have a significant affects on the maximum rate that material can be fed and possibly more importantly, how well separation is achieved. Also, origin, cracking and type of cocoa all affect the final yield and separation. 137 lbs of beans (100 lbs of nibs) and 3 hours later I can make these observations. Panama is a dream to winnow. It cracks very evenly. The husk is small and I was able to introduce the nib/husk mixture at a rate of nearly 2 lbs/min with 75.6% yield and virtually no visable husk in the nibs.
Ghana is pretty good, but the heavier husk of the Forastero like to stay in larger pieces. Separation was still high at 75.5%, but I could see about 0.5 to 1% husk in the nibs.
Madagascar was difficult. It is a drier than average bean, so powdered a little more than I expected. Consequently, the separation was ok with very little visible husk in the nibs, but yield was down to 69%.
Next on the drawing board is the feed mechanism. It's technically drawn up, but I need to build it next. After that, I want to experiment a little with the primary chamber design shape. I was noticing the beginning of some nice circular flow patterns and I am curious is modifying the chamber shape to encourage these might allow me to further increase my feed rate while maintaining a high yield.